# Fano 3-folds in Open image in new window format, Tom and Jerry

## Abstract

We study \({\mathbb {Q}}\)-factorial terminal Fano 3-folds whose equations are modelled on those of the Segre embedding of Open image in new window . These lie in codimension 4 in their total anticanonical embedding and have Picard rank 2. They fit into the current state of classification in three different ways. Some families arise as unprojections of degenerations of complete intersections, where the generic unprojection is a known prime Fano 3-fold in codimension 3; these are new, and an analysis of their Gorenstein projections reveals yet other new families. Others represent the “second Tom” unprojection families already known in codimension 4, and we show that every such family contains one of our models. Yet others have no easy Gorenstein projection analysis at all, so prove the existence of Fano components on their Hilbert scheme.

## Keywords

Fano 3-fold Segre embedding Gorenstein format## Mathematics Subject Classification

14J30 14J10 14J45 14M07## 1 Introduction

### 1.1 Fano 3-folds, Gorenstein rings and Open image in new window

*Fano 3-fold*is a complex projective variety

*X*of dimension 3 with \({\mathbb {Q}}\)-factorial terminal singularities and \(-K_X\) ample. We construct several new Fano 3-folds, and others which explain known phenomena. The anticanonical ringof a Fano 3-fold

*X*is Gorenstein, and provides an embedding \(X\subset \mathrm{w}\mathbb P\) in weighted projective space (wps) that we exploit here, focusing on the case \(X\subset \mathrm{w}\mathbb P^7\) of codimension 4.

The number of deformation families of Fano 3-folds is finite [20, 21], and the Graded Ring Database (Grdb) [4, 6] has a list of rational functions Open image in new window that includes all Hilbert series Open image in new window of Fano 3-folds with Open image in new window . (In fact, we do not know of any Fano 3-fold whose Hilbert series is not on that list, even without this additional condition.) An attempt at an explicit classification, outlined in [2], aims to describe all deformation families of Fano 3-folds for each such Hilbert series. All families whose general member lies in codimension \(\leqslant 2\) are known [12], and almost certainly those in codimension 3 are too [2, 6]. An analysis of (Gorenstein) projections [8, 24, 34] provides much of the classification in codimension 4, but it is not complete, and codimension 4 remains at the cutting edge.

We use the methods of [8] freely, although we work through an example in detail in Sect. 3 and explain any novelties as they arise.

### 1.2 The aims of this paper

We describe families of Fano 3-folds \(X\subset \mathrm{w}\mathbb P^7\) whose equations are a specialisation of the format (1); that is, they are regular pullbacks, as in Sect. 2. It is usually hard to describe the equations of varieties in codimension 4—see papers from Kustin and Miller [22] to Reid [31]—but if we decree the format in advance, then the equations come almost for free, and the question becomes how to put a grading on them to give Fano 3-folds. Our results come in three broad flavours, which we explain in Sects. 4–6 and summarise here.

**Section** 4: **Unprojecting Pfaffian degenerations.** We find new varieties in Open image in new window format that have the same Hilbert series as known Fano 3-folds but lie in different deformation families. From another point of view, we understand this as the unprojection analysis of degenerations of complete intersections, and this treatment provides yet more families not exhibited by [8]. (The key point is that the unprojection divisor \(D\subset Y\) does not persist throughout the degeneration \(Y\leadsto Y_0\), and so the resulting unprojection is not a degeneration in a known family.)

### Theorem 1.1

There are three deformation families of Fano 3-folds *X* with Hilbert series Open image in new window . Their respective general members Open image in new window all lie in codimension 4 with degree \(-K_X^3 = 17/2\) and a single orbifold singularity \(\frac{1}{2}(1,1,1)\), and with invariants:

We prove this particular result in Sect. 3; the last two columns of the table refer to the unprojection calculation (*N* is the number of nodes, as described in Sect. 3), which is explained in the indicated sections. The Euler characteristic *e*(*X*) is calculated during the unprojection following [8, Section 7] and the other invariants follow. We do not know whether there are any other deformation families realising the same Hilbert series Open image in new window .

*X*by \(A_X\), [15, Theorem 2.5] gives

In this case, all three families lie in codimension 4. It is more common that the known family lies in codimension 3 and we find new families in codimension 4. Thus the corresponding Hilbert scheme contains different components whose general members are Fano 3-folds in different codimensions, a phenomenon we had not seen before.

Further analysis of degenerations finds yet more new Fano 3-folds even where there is no Open image in new window model; the following result is proved in Sect. 4.2.

### Theorem 1.2

There are two deformation families of Fano 3-folds *X* with Hilbert series \(P_X = P_{548}\). Their respective general members *X* have degree \(-K_X^3 = 1/15\), and are distinguished by their embedding in wps and Euler characteristics as follows:

\(X\subset \mathrm{w}\mathbb P\) | | \(\#\,\mathrm{nodes}\) | |
---|---|---|---|

Family 1 | \(-\,42\) | 8 | |

Family 2 | \(-\,40\) | 9 |

In this case there is no Open image in new window model: such a model would come from a specialised Tom unprojection, but the Tom and Jerry analysis outlined in Sect. 4.2 rules this out.

**Section** 5: **Second Tom.** The Big Table [9] lists all (general) Fano 3-folds in codimension 4 that have a Type I projection. Such projections can be of Tom type or Jerry type (see [8, 2.3]). The result of that paper is that every Fano 3-fold admitting a Type I projection has at least one Tom family and one Jerry family. However in some cases there is a second Tom or second Jerry (or both). Two of these cases were already known to Szendrői [32], even before the Tom and Jerry analysis was developed.

Euler characteristic is of course constant in families, but whenever there is a second Tom, the Euler characteristics of members of the two Tom families differ by 2. Theorem 5.1 below says that in this case the Tom family with smaller Euler characteristic always contains special members in Open image in new window format.

**Section**6:

**No Type I projection.**Finally, we find some Fano 3-folds that are harder to describe, including some that currently have no construction by Gorenstein unprojection. Such Fano 3-folds were expected to exist, but this is the first construction of them in the literature we are aware of. It may be the case that there are other families of such Fano 3-folds having Picard rank 1, but our methods here cannot answer that question.

53 Fano 3-fold Hilbert series in Open image in new window format (number of nodes is given as a superscript to \(\mathrm{Tom}\)/\(\mathrm{Jer}\))

| | | Grdb | | T/J | \(\mathrm{w}\mathbb P\) in Grdb | codim 4 models in this paper |
---|---|---|---|---|---|---|---|

4 | 000 | 112 | 26989 | 4 | \(\mathrm{Tom}^5\!, \mathrm{Jer}^7\) in Open image in new window | ||

5 | 000 | 122 | 20652 | 4 | TTJ | Second Tom | |

5 | 001 | 112 | 20543 | 3 | n/a | \(\mathrm{Tom}^7\!, \mathrm{Jer}^9\) in Open image in new window | |

5 | 001 | 112 | 24078 | 4 | TTJ | Second Tom | |

6 | 000 | 222 | 12960 | 4 | TJ | Subfamily of Tom | |

6 | 001 | 122 | 16339 | 4 | TTJJ | Second Tom | |

7 | 001 | 123 | 11436 | 3 | n/a | \(\mathrm{Tom}^{13}\) in Open image in new window | |

7 | 001 | 123 | 16228 | 4 | TTJJ | Second Tom | |

7 | 011 | 122 | 11455 | 4 | TTJJ | Second Tom | |

8 | 001 | 223 | 11157 | 5 | n/a | Bad 1 / 4 point | |

8 | 001 | 223 | 6878 | 4 | TTJJ | Second Tom | |

8 | 011 | 123 | 11125 | 4 | TTJJ | Second Tom | |

9 | 001 | 233 | 5970 | 4 | TTJJ | Second Tom | |

9 | 012 | 123 | 11106 | 4 | TTJJ | Second Tom | |

9 | 012 | 123 | 11021 | 4 | TTJJ | Second Tom | |

9 | 012 | 123 | 5962 | 3 | n/a | \(\mathrm{Tom}^{11}\!,\mathrm{Jer}^{13}\) in Open image in new window | |

9 | 012 | 123 | 6860 | 4 | TTJ | Second Tom | |

10 | 001 | 234 | 5870 | 4 | TTJJ | Second Tom | |

10 | 011 | 233 | 5530 | 4 | TTJJ | Second Tom | |

10 | 012 | 124 | 10984 | 3 | n/a | Bad 1/4 point | |

10 | 012 | 124 | 5858 | 3 | n/a | \(\mathrm{Tom}^{13}\!,\mathrm{Jer}^{14}\) in Open image in new window | |

11 | 011 | 234 | 5306 | 4 | TTJJ | Second Tom | |

11 | 012 | 134 | 5302 | 3 | n/a | \(\mathrm{Tom}^{16}\) in Open image in new window | |

11 | 012 | 134 | 5844 | 3 | n/a | Bad 1 / 6 point and no 1 / 5 | |

11 | 012 | 134 | 10985 | 4 | TTJJ | Second Tom | |

12 | 012 | 234 | 1766 | 4 | no I | Quasismooth Open image in new window model | |

12 | 012 | 234 | 5215 | 4 | TTJJ | Second Tom | |

12 | 012 | 234 | 2427 | 4 | TTJJ | Second Tom | |

12 | 012 | 234 | 5268 | 4 | TTJJ | Second Tom | |

13 | 001 | 345 | 1413 | 4 | TTJJ | Second Tom | |

13 | 012 | 235 | 5177 | 4 | TJ | Bad 1/5 point | |

13 | 012 | 235 | 2422 | 4 | TTJJ | Second Tom | |

14 | 011 | 345 | 5002 | 4 | TTJJ | Second Tom | |

14 | 012 | 245 | 5163 | 4 | TTJJ | Second Tom | |

14 | 012 | 245 | 1410 | 4 | TJJ | Bad 1/4 point | |

14 | 013 | 235 | 4999 | 3 | n/a | Bad 1/4 point | |

14 | 013 | 235 | 1396 | 3 | n/a | \(\mathrm{Tom}^9\!,\mathrm{Jer}^{11}\) in Open image in new window | |

15 | 012 | 345 | 878 | 4 | no I | Quasismooth Open image in new window model | |

15 | 012 | 345 | 4949 | 4 | TTJJ | Second Tom | |

15 | 012 | 345 | 1253 | 4 | TTJ | Second Tom | |

15 | 012 | 345 | 1218 | 4 | TTJJ | Second Tom | |

15 | 012 | 345 | 4989 | 4 | TTJJ | Second Tom | |

16 | 012 | 346 | 1186 | 4 | TJJ | Bad 1/5 point | |

17 | 012 | 356 | 648 | 4 | no I | Bad 1/5 point | |

17 | 012 | 356 | 4915 | 4 | TTJJ | Second Tom | |

18 | 012 | 456 | 577 | 4 | no I | Quasismooth but not terminal | |

18 | 012 | 456 | 645 | 4 | TJ | Bad 1 / 4 point | |

18 | 012 | 456 | 4860 | 4 | TTJJ | Second Tom | |

19 | 012 | 457 | 570 | 4 | TJJ | Bad 1/5 point | |

20 | 012 | 467 | 4839 | 4 | TTJJ | Second Tom | |

22 | 012 | 568 | 1091 | 4 | TJJ | Bad 1/7 point | |

22 | 012 | 568 | 393 | 4 | TJ | Bad 1/4, 1/5 points | |

23 | 012 | 578 | 360 | 4 | no I | Bad 1/7 point |

### 1.3 Summary of results

Our approach starts with a systematic enumeration of all possible Open image in new window formats that could realise the Hilbert series of a Fano 3-fold after appropriate specialisation. In Sect. 7, following [7, 27], we find 53 varieties in Open image in new window format that have the Hilbert series of a Fano 3-fold. We summarise the fate of each of these 53 cases in Table 1; the final column summarises our results, as we describe below, and the rest of the paper explains the calculations that provide the proof.

The columns of Table 1 are as follows. Column *k* is an adjunction index, described in Sect. 7.1, and columns *a* and *b* refer to the vectors in Sect. 2 that determine the weights on the weighted Open image in new window . Column Grdb lists the number of the Hilbert series in [6], column *c* indicates the codimension of the usual model suggested there, and \(\mathrm{w}\mathbb P\) its ambient space. Column T/J shows the number of distinct Tom and Jerry components according to [8]. For example, TTJ indicates there are two Tom unprojections and one Jerry unprojection in the Big Table [9]. We write ‘no I’ when the Hilbert series does not admit a numerical Type I projection, and so the Tom and Jerry analysis does not apply, and ‘n/a’ if the usual model is in codimension 3 rather than 4.

The final column describes the results of this paper; it is an abbreviation of more detailed results. For example, Theorem 1.1 expands out the first line of the table, \(k=4\), and other lines of the table that are not indicated as failing have analogous theorems that the final column summarises. If the Open image in new window model fails to realise a Fano 3-fold at all, it is usually because the general member does not have terminal singularities; we say, for example, ‘bad 1/4 point’ if the format forces a non-quasismooth, non-terminal index 4 point onto the variety.

When the Grdb model is in codimension 3, we list which Tom and Jerry unprojections of a degeneration work to give alternative varieties in codimension 4, indicating the number of nodes as a superscript and the codimension 4 ambient space. (We do not say which Tom or Jerry since that depends on a choice of rows and columns.) In each case the Tom unprojection gives the Open image in new window model determined by the parameters *a* and *b*. The usual codimension 3 model arises by Type I unprojection with number of nodes being one more that that of the Open image in new window Tom model.

When the Grdb model is in codimension 4 with two Tom unprojections, the Open image in new window always works to give the second of the Tom families. The further Tom and Jerry analysis of the unprojection is carried out in [8] and we do not repeat the result here. When the Grdb model is in codimension 4 with only a single Tom unprojection, the model usually fails. The exception is family 12,960, which does work as a Open image in new window model. There is also a case of a Hilbert series, number 11,157, where the Grdb offers a prediction of a variety in codimension 5, but this fails as a Open image in new window model.

In Sect. 7.1, we outline a computer search that provides the *a*, *b* parameters of Table 1 which are the starting point of the analysis here. In Sect. 7.2, we summarise the results of [32] that provide the most general form of the Hilbert series of a variety in Open image in new window format; that paper also discovered cases 11,106 and 11,021 of Table 1 that inspired our approach here. First we introduce the key varieties of the Open image in new window format in Sect. 2.

## 2 The key varieties and weighted Open image in new window formats

The affine cone Open image in new window on Open image in new window is defined by the equations (1) on \({\mathbb {C}}^9\). It admits a 6-dimensional family of \({\mathbb {C}}^*\) actions, or equivalently six degrees of freedom in assigning positive integer gradings to its (affine) coordinate ring. We express this as follows.

*weighted*Open image in new window aswhere the variables have weightsThus Open image in new window , where the \({\mathbb {C}}^*\) action is determined by the grading. We treat Open image in new window as a key variety for each different pair

*a*,

*b*. (Note that the entries of

*a*and

*b*may also all lie in \(\frac{1}{2}+\mathbb Z\), without any change to our treatment here).

### Proposition 2.1

Open image in new window is a 4-dimensional, \({\mathbb {Q}}\)-factorial projective toric variety of Picard rank \(\rho _V = 2\).

### Proof

First we describe a toric variety Open image in new window by its Cox ring. The input data is the weight matrix (3), which is weakly increasing along rows and down columns. The key is to understand the freedom one has to choose alternative vectors \(a^{(i)}\!,b^{(i)}\), for \(i = 1,2\), to give the same matrix. For example, if we choose \(a^{(1)}_1\!=0\), then \(b^{(1)}\) is determined by the top row, and then \(a^{(1)}_2\) and \(a^{(1)}_3\) are determined by the first column. Alternatively, choosing \(b^{(2)}_1\!=0\) determines different vectors \(a^{(2)}\) and \(b^{(2)}\). Concatenating the *a* and *b* vectors to give Open image in new window determines a 2-dimensional \({\mathbb {Q}}\)-subspace \(U=U_{a,b}\subset {\mathbb {Q}}^6\) together with a chosen integral basis .

*R*in variables \(u_1,u_2,u_3\), \(v_1,v_2,v_3\), bi-graded by the columns of the matrix (giving the two \({\mathbb {C}}^*\) actions)

*U*). The bilinear mapis an isomorphism onto its image Open image in new window , and the conclusions of the proposition all follow at once. (\({\mathbb {Q}}\)-factoriality holds since the Cox coordinates correspond to the 1-skeleton of the fan, and so any maximal cone with at least five rays must contain all \(u_i\) or all Open image in new window , contradicting the choice of irrelevant ideal.)

If Open image in new window is not well formed, then, just as for wps, there is a different weight matrix that is well formed and determines a toric variety \(W'\) isomorphic to Open image in new window . (See Iano-Fletcher [18, 6.9–20] for wps and Ahmadinezhad [1, 2.3] for the general case.) The proposition follows using \(W'\). \(\square \)

The well forming process used in the proof is easy to use. For example, if an integer \(n>1\) divides every entry of some row of the weight matrix (4), then we may divide that row through by *n*; the subspace \(U\subset {\mathbb {Q}}^6\) is unchanged by this. Or if an integer \(n>1\) divides all columns except one, then the corresponding Cox coordinate *u* appears only as \(u^n\) in the coordinate rings of standard affine patches and we may truncate *R* by replacing the generator *u* by \(u^n\); this does not change the coordinate rings of the affine patches, and so the scheme it defines is isomorphic to the original (c.f. [1, Lemma 2.9] for the more general statement). This multiplies the *u* column of (4) by *n*, changing the subspace \(U\subset {\mathbb {Q}}^6\), and then we may divide the whole matrix by *n* as before. See [1, 2.3] for the complete process.

Having said that, in practice we will work with non-well-formed quotients if they arise, since they still admit regular pullbacks that are well formed, and the grading on the target wps is something we fix in advance. More importantly for us here is that well forming step \(u\leadsto u^n\) destroys the Open image in new window structure, so we avoid it.

### Example 2.2

*W*, the Segre map is not bi-linear: \(u_1v_1\) has bidegree \(\left( {\begin{matrix} 1\\ 1 \end{matrix}}\right) \), but \(u_1v_3\) has an independent bidegree \(\left( {\begin{matrix} 3\\ 2 \end{matrix}}\right) \). We could use \(u_1^2v_3\) instead, which has proportional bidegree \(\left( {\begin{matrix} 3\\ 3 \end{matrix}}\right) \). Taking Open image in new window , where

*R*is the graded ring of forms of degrees \(\left( {\begin{matrix} m\\ m \end{matrix}}\right) \) for \(m\geqslant 0\), gives Open image in new window , which is now well formed, but we have lost the codimension 4 property of

*V*we want to exploit. In a case like this, we work directly with the non-well-formed Open image in new window and its non-well-formed image Open image in new window .

We use the varieties Open image in new window as key varieties to produce new varieties from by regular pullback; see [30, Section 1.5] or [7, Section 2]. In practical terms, that means writing equations in the form of (1) inside a wps \(\mathrm{w}\mathbb P^7\) where the \(x_i\) are homogeneous forms of positive degrees, and the resulting loci \(X\subset \mathrm{w}\mathbb P^7\) are the Fano 3-folds we seek.

Alternatively, we may treat *X* as a complete intersection in a projective cone over Open image in new window , as in Sect. 3.2 below, where the additional cone vertex variables may have any positive degrees; this point of view is taken by Corti-Reid and Szendrői in [14, 26, 29, 32]. It follows from this description that the Picard rank of *X* is 2.

## 3 Unprojection and the proof of Theorem 1.1

*Y*is quasismooth away from

*N*nodes, all of which lie on

*D*, then

### 3.1 The classical Open image in new window family

*Y*has six nodes that lie on

*D*: in coordinates

*x*,

*y*,

*z*,

*u*,

*v*,

*w*,

*t*of Open image in new window , setting Open image in new window , the general

*Y*has equations defined by

*X*has a Open image in new window free resolution. If \(Y_{\mathrm {gen}}\) is a nonsingular small deformation of

*Y*, then Open image in new window (by the usual Chern class calculation, since \(Y_{\mathrm {gen}}\) is a smooth 2, 2, 2 complete intersection) so, by (6),This family is described by Takagi [33]; it is no. 1.4 in the tables there of Fano 3-folds of Picard rank 1.

### 3.2 A Open image in new window family with Tom projection

Consider the Open image in new window key variety Open image in new window , where \(a = \bigl (\frac{1}{2},\frac{1}{2},\frac{1}{2}\bigr )\) and \(b=\bigl (\frac{1}{2},\frac{1}{2},\frac{3}{2}\bigr )\). We define a quasismooth variety Open image in new window in codimension 4 as a regular pullback.

*x*,

*y*,

*z*,

*t*,

*u*,

*v*,

*w*,

*s*on Open image in new window , a Open image in new window matrix

*M*of forms of degrees

*y*,

*z*,

*u*,

*v*are implicit functions in a neighbourhood of \(P_s\in X^{(2)}\). The Gorenstein projection from this point \(P_s\) has image Open image in new window , where

*N*. (The nonzero entries of

*N*are those of \(M^T\) with the entry

*s*deleted.)

This *Y* contains the projection divisor Open image in new window and has five nodes on *D* (either by direct calculation, or by the formula of [8, Section 7]). The divisor \(D\subset Y\) is in Tom\(_3\) configuration: entries \(n_{i,j}\) of the skew Open image in new window matrix *N* defining *Y* lie in the ideal \(I_D = (y,z,u,v)\) if both \(i\not =3\) and \(j\not =3\); that is, all entries off row 3 and column 3 of *N* are in \(I_D\). Thus, in particular, we can reconstruct \(X^{(2)}\) from \(D\subset Y\) as the Tom\(_3\) unprojection. It follows from Papadakis–Reid [25, Section 2.4] that \(\omega _{X^{(2)}} = \mathscr {O}_{X^{(2)}}(-1)\) and so \(X^{(2)}\) is a Fano 3-fold.

It remains to show that Open image in new window , so that this Fano 3-fold must lie in a different deformation family from the classical one constructed in Sect. 3.1.

*N*is in fact zero while the degree of \(f_{4,5}\) is 2, although each entry is of course the zero polynomial in this case; we denote this by indicating the degrees of the entries with brackets around those that are zero in this case:

*Y*by varying these two entries to Open image in new window and Open image in new window , where \(\varepsilon \not =0\) and

*f*is a general quadric on Open image in new window (and, of course, the skew symmetric entries in \(f_{2,1}\) and \(f_{5,4}\)). Denoting the deformed matrix by \(N_\varepsilon \), and Open image in new window , we see a small deformation of

*Y*to a smooth Fano 3-fold Open image in new window that is a 2, 2, 2 complete intersection. (The nonzero constant entries of \(N_\varepsilon \) provide two syzygies that eliminate two of the five Pfaffians.) As in Sect. 3.1, the smoothing \(Y_\varepsilon \) has Euler characteristic Open image in new window , so by (6) we have that Open image in new window .

Note that the Pfaffian smoothing \(Y_\varepsilon \) of *Y* destroys the unprojection divisor \(D\subset Y\): for *D* to lie inside \(Y_\varepsilon \) the entries \(f_{3,4}\) and \(f_{3,5}\) of \(N_\varepsilon \) would have to lie in \(I_D\) (so \(N_\varepsilon \) would be in Jer\(_{4,5}\) format with the extra constraint \(f_{4,5}=0\)), but then *Y* would be singular along *D* since three of the five Pfaffians would lie in \(I_D^2\).

### 3.3 A third family by Jerry unprojection

*N*with weights

*N*lying in \(I_D\) whenever

*i*or

*j*lie in \(\{1,3\}\). The general such \(D\subset Y\) has seven nodes on

*D*. Unprojecting \(D\subset Y\) gives a general member \(X^{(3)}\) of a third family with Open image in new window .

This completes the proof of Theorem 1.1.

## 4 Unprojecting Pfaffian degenerations

### 4.1 Open image in new window models with a codimension 3 Pfaffian component

Each of the Fano Hilbert series 1396, 5302, 5858, 5962, 11436, 20543 is realised by a codimension 3 Pfaffian model, which is the simple default model presented in the Grdb. (So too are 4999, 5844 and 10,984, but we do not find new models for these.) We show that they can also be realised by a Open image in new window model in a different deformation family (and sometimes a third model too). The key point is that a projection of the usual model admits alternative degenerations in higher codimension that also contain a divisor that can be unprojected.

*M*has degrees

*X*as a Type I unprojection ofIn general,

*Y*has eight nodes lying on

*D*, and it smooths to a nonsingular Fano 3-fold \(Y_{\mathrm {gen}}\) with Euler characteristic Open image in new window . Thus a general

*X*has Euler characteristic Open image in new window .

**A quasismooth**Open image in new window

**family.**We can write another (quasismooth) model Open image in new window in codimension 4 in Open image in new window format with weights

*M*has degrees

*Y*has seven nodes lying on

*D*; in coordinates

*x*,

*y*,

*z*,

*t*,

*u*,

*w*,

*v*, we may take \(D=\mathbb P^2\) to be Open image in new window . By varying the (1, 2) entry from zero to a unit,

*Y*has a deformation to a quasismooth 3, 3 complete intersection \(Y_{\mathrm {gen}}\) as before, and so, Open image in new window . Thus these Open image in new window models are members of a different deformation family from the original one.

More is true in this case: the general member of this new deformation family is in Open image in new window format. Starting with matrix (7) and \(D=\mathbb P^2\) as above, the (1, 2) entry of the general Tom\(_3\) matrix is necessarily the zero polynomial. In general, the four entries (1, 4), (1, 5), (2, 4) and (2, 5) of the matrix are in the ideal Open image in new window , and for the general member these four variables are dependent on those entries. Thus the (4, 5) entry can be arranged to be zero by row-and-column operations.

**Another family in codimension 4.** There is a third deformation family in this case. The codimension 3 format (7) also admits a Jerry\(_{15}\) unprojection with nine nodes on *D*, giving Open image in new window in codimension 4 with Open image in new window .

### 4.2 Pfaffian degenerations of codimension 2 Fano 3-folds

The key to the cases in Sect. 4.1 that the Open image in new window model exposes is the degeneration of a codimension 2 Fano 3-fold. More generally, Table 3 of [3] lists 13 cases of Fano 3-fold degenerations where the generic fibre is a codimension 2 complete intersection and the special fibre is a codimension 3 Pfaffian. In each case, the anti-symmetric Open image in new window syzygy matrix of the special fibre has an entry of degree 0, which is the zero polynomial in the degeneration, but when nonzero serves to eliminate a single variable. (In fact [3] describes the graded rings of K3 surfaces, but these extend to Fano 3-folds by the usual extension–deformation method introducing a new variable of degree 1.)

*Y*has eight nodes; the unprojection of \(D\subset Y\) gives the codimension 3 Pfaffian familyImposing the same unprojection divisor \(D\subset Y^{0}\) can be done in two distinct ways, coming from different Tom and Jerry arrangments. In one way, there are degenerations \(Y^{t}_{12,13} \leadsto Y^{0}\) which contain the same

*D*in every fibre \(Y_t\). These unproject to a degeneration of the family (8) by the following lemma: indeed unprojection commutes with regular sequences by [10, Lemma 5.6], and so unprojection commutes with flat deformation, if one fixes the unprojection divisor; so the lemma is a particular case of [10, Lemma 5.6].

### Lemma 4.1

Let Open image in new window be any wps and fix Open image in new window , for some \(d\leqslant s-2\). Suppose \(Y_t\subset \mathscr {Y}\rightarrow \mathscr {T}\) is a flat 1-dimensional family of projectively Gorenstein subschemes of \(\mathbb P\) over smooth base \(0\in \mathscr {T}\), each one containing *D* and with \(\dim Y_t = \dim D + 1=d+1\), and with \(\omega _Y=\mathscr {O}_Y(k_Y)\). Let Open image in new window be the unprojection of Open image in new window , where \(b = k_Y-k_D = a_0+\cdots +a_d-1\). Then \(\mathscr {X}\) is flat over \(\mathscr {T}\), and for each closed point \(t\in \mathscr {T}\) the fibre \(X_t\in \mathscr {X}\) is the unprojection of \(D\subset Y_t\).

*D*. Indeed, in this \(D\subset Y^0\) model, \(Y^{0}\) has nine nodes on

*D*, which is a numerical obstruction to any such deformation. This \(D\subset Y^{0}\) unprojects to a codimension 4 Fano 3-foldwith the same Hilbert Series No. 548 as (8) but lying in a different component: it has Euler characteristic Open image in new window . This proves Theorem 1.2.

## 5 Open image in new window and the second Tom

The Big Table [9], which contains the results of [8], lists deformation families of Fano 3-folds in codimension 4 that have a Type I projection to a Pfaffian 3-fold in codimension 3. The components are listed according to the Tom or Jerry type of the projection: the type of projection is invariant for sufficiently general members of each component. The result of this section gives an interpretation of the Big Table of [8], but does not describe any new families of Fano 3-folds.

### Theorem 5.1

For every Hilbert series listed in the Big Table [9] that is realised by two distinct Tom projections, there is a Fano 3-fold in Open image in new window format that lies on the family containing 3-folds with the smaller (more negative) Euler characteristic.

Hilbert series in Open image in new window format that admit a second Tom unprojection

Grdb | Open image in new window weights | T/J families | Centre: # nodes |
---|---|---|---|

1253 | \(\left( \begin{array}{ccc} 3&{}4&{}5\\ 4&{}5&{}6\\ 5&{}6&{}7 \end{array}\right) \) | TTJ | \(\frac{1}{7}{:}6\) |

1218 | \(\left( \begin{array}{ccc} 3&{}4&{}5\\ 4&{}5&{}6\\ 5&{}6&{}7 \end{array}\right) \) | TTJJ | \(\frac{1}{5}{:}9\) |

1413 | \(\left( \begin{array}{ccc} 3&{}4&{}5\\ 3&{}4&{}5\\ 4&{}5&{}6 \end{array}\right) \) | TTJJ | \(\frac{1}{5}{:}7\) |

2422 | \(\left( \begin{array}{ccc} 2&{}3&{}5\\ 3&{}4&{}6\\ 4&{}5&{}7 \end{array}\right) \) | TTJJ | \(\frac{1}{7}{:}5\) |

2427 | \(\left( \begin{array}{ccc} 2&{}3&{}4\\ 3&{}4&{}5\\ 4&{}5&{}6 \end{array}\right) \) | TTJJ | \(\frac{1}{5}{:}6\) |

4839 | \(\left( \begin{array}{ccc} 4&{}6&{}7\\ 5&{}7&{}8\\ 6&{}8&{}9 \end{array}\right) \) | TTJJ | \(\frac{1}{5}{:}20\); \(\frac{1}{9} {:} 13\) |

4860 | \(\left( \begin{array}{ccc} 4&{}5&{}6\\ 5&{}6&{}7\\ 6&{}7&{}8 \end{array}\right) \) | TTJJ | \(\frac{1}{7}{:}13\) |

4915 | \(\left( \begin{array}{ccc} 3&{}5&{}6\\ 4&{}6&{}7\\ 5&{}7&{}8 \end{array}\right) \) | TTJJ | \(\frac{1}{4}{:}19\); \(\frac{1}{8}{:}11\) |

4949 | \(\left( \begin{array}{ccc} 3&{}4&{}5\\ 4&{}5&{}6\\ 5&{}6&{}7 \end{array}\right) \) | TTJJ | \(\frac{1}{6}{:}11\) |

4989 | \(\left( \begin{array}{ccc} 3&{}4&{}5\\ 4&{}5&{}6\\ 5&{}6&{}7 \end{array}\right) \) | TTJJ | \(\frac{1}{4}{:}15\); \(\frac{1}{7}{:}10\) |

5002 | \(\left( \begin{array}{ccc} 3&{}4&{}5\\ 4&{}5&{}6\\ 4&{}5&{}6 \end{array}\right) \) | TTJJ | \(\frac{1}{4}{:}14\); \(\frac{1}{5}{:}11\); \(\frac{1}{6}{:}10\) |

5163 | \(\left( \begin{array}{ccc} 2&{}4&{}5\\ 3&{}5&{}6\\ 4&{}6&{}7 \end{array}\right) \) | TTJJ | \(\frac{1}{3}{:}19\); \(\frac{1}{7}{:}9\) |

5215 | \(\left( \begin{array}{ccc} 2&{}3&{}4\\ 3&{}4&{}5\\ 4&{}5&{}6 \end{array}\right) \) | TTJJ | \(\frac{1}{5}{:} 9\) |

5268 | \(\left( \begin{array}{ccc} 2&{}3&{}4\\ 3&{}4&{}5\\ 4&{}5&{}6 \end{array}\right) \) | TTJJ | \(\frac{1}{3}{:}14\); \(\frac{1}{5}{:}8\) |

5306 | \(\left( \begin{array}{ccc} 2&{}3&{}4\\ 3&{}4&{}5\\ 3&{}4&{}5 \end{array}\right) \) | TTJJ | \(\frac{1}{3}{:}13\); \(\frac{1}{4}{:}9\); \(\frac{1}{5}{:}8\) |

5530 | \(\left( \begin{array}{ccc} 2&{}3&{}3\\ 3&{}4&{}4\\ 3&{}4&{}4 \end{array}\right) \) | TTJJ | \(\frac{1}{3}{:}11\); \(\frac{1}{4}{:}8\) |

5870 | \(\left( \begin{array}{ccc} 2&{}3&{}4\\ 2&{}3&{}4\\ 3&{}4&{}5 \end{array}\right) \) | TTJJ | \(\frac{1}{3}{:}10\); \(\frac{1}{5}{:}7\) |

5970 | \(\left( \begin{array}{ccc} 2&{}3&{}3\\ 2&{}3&{}3\\ 3&{}4&{}4 \end{array}\right) \) | TTJJ | \(\frac{1}{3}{:}9\); \(\frac{1}{4}{:} 7\) |

6860 | \(\left( \begin{array}{ccc} 1&{}2&{}3\\ 2&{}3&{}4\\ 3&{}4&{}5 \end{array}\right) \) | TTJ | \(\frac{1}{5}{:}4\) |

6878 | \(\left( \begin{array}{ccc} 2&{}2&{}3\\ 2&{}2&{}3\\ 3&{}3&{}4 \end{array}\right) \) | TTJJ | \(\frac{1}{3}{:}8\) |

10985 | \(\left( \begin{array}{ccc} 1&{}3&{}4\\ 2&{}4&{}5\\ 3&{}5&{}6 \end{array}\right) \) | TTJJ | \(\frac{1}{2}{:}23\); \(\frac{1}{6}{:}7\) |

11021 | \(\left( \begin{array}{ccc} 1&{}2&{}3\\ 2&{}3&{}4\\ 3&{}4&{}5 \end{array}\right) \) | TTJJ | \(\frac{1}{4}{:}7\) |

11106 | \(\left( \begin{array}{ccc} 1&{}2&{}3\\ 2&{}3&{}4\\ 3&{}4&{}5 \end{array}\right) \) | TTJJ | \(\frac{1}{2}{:}15\); \(\frac{1}{5}{:}6\) |

11125 | \(\left( \begin{array}{ccc} 1&{}2&{}3\\ 2&{}3&{}4\\ 2&{}3&{}4 \end{array}\right) \) | TTJJ | \(\frac{1}{2}{:}14\); \(\frac{1}{3}{:}7\); \(\frac{1}{4}{:}6\) |

11455 | \(\left( \begin{array}{ccc} 1&{}2&{}2\\ 2&{}3&{}3\\ 2&{}3&{}3 \end{array}\right) \) | TTJJ | \(\frac{1}{2}{:}11\); \(\frac{1}{3}{:}6\) |

16228 | \(\left( \begin{array}{ccc} 1&{}2&{}3\\ 1&{}2&{}3\\ 2&{}3&{}4 \end{array}\right) \) | TTJJ | \(\frac{1}{2}{:}9\); \(\frac{1}{4}{:}5\) |

16339 | \(\left( \begin{array}{ccc} 1&{}2&{}2\\ 1&{}2&{}2\\ 2&{}3&{}3 \end{array}\right) \) | TTJJ | \(\frac{1}{2}{:}8\); \(\frac{1}{3}{:}5\) |

20652 | \(\left( \begin{array}{ccc} 1&{}2&{}2\\ 1&{}2&{}2\\ 1&{}2&{}2 \end{array}\right) \) | TTJ | \(\frac{1}{2}{:}6\) |

24078 | \(\left( \begin{array}{ccc} 1&{}1&{}2\\ 1&{}1&{}2\\ 2&{}2&{}3 \end{array}\right) \) | TTJ | \(\frac{1}{3}{:}4\) |

*X*has Type I projections from both \(\frac{1}{5}(1,1,4)\) and \(\frac{1}{9}(1,1,8)\). (It is enough to consider just one of these centres of projection, but [8] calculates both, drawing the same conclusion twice.)

*x*,

*y*,

*z*,

*t*,

*u*,

*v*,

*w*,

*s*. The Open image in new window minors of the matrix

Eliminating either the variable *t* of degree 5 or *s* of degree 9 computes the two possible Type I projections, with image a nodal codimension 3 Fano 3-fold *Y* containing Open image in new window or \({D=\mathbb P(1,1,8)}\) with 20 or 13 nodes lying on *D* respectively. (Both *t* and *s* appear only once in the matrix, so eliminating them simply involves omitting that entry and mounting the rest of the matrix in a skew matrix, as usual.)

## 6 Cases with no numerical Type I projection

The five Hilbert series 360, 577, 648, 878 and 1766 do not admit a Type I projection, and so the analysis of [8] does not apply. Nevertheless each is realised by a variety in Open image in new window format exist, although only two of these are Fano 3-folds.

In the two cases 360 and 648 the general Open image in new window model is not quasismooth and has a non-terminal singularity, so there is no Open image in new window Fano model. (Each of these admit Type II\(_1\) projections, so are instead subject to the analysis of [24]; this is carried out by Taylor [34].) In the case 577, the Open image in new window model is quasismooth, but it has a \(\frac{1}{4}(1,1,1)\) quotient singularity and so is not a terminal Fano 3-fold and again there is no Open image in new window Fano model.

## 7 Enumerating Open image in new window formats

### 7.1 Enumerating Open image in new window formats and cases that fail

*R*(

*X*) satisfy the orbifold integral plurigenus formula [11, Theorem 1.3]where \(P_{\mathrm {ini}}\) is a function only of the genus

*g*of

*X*, where \(g+2=h^0(-K_X)\), and \(P_{\mathrm {orb}}\) is a function of a quotient singularity \(Q = \frac{1}{r}(1,a,-a)\), the collection of which form the basket \(\mathscr {B}\) of

*X*(see [13, Section 9]). When \(X\subset \mathrm{w}\mathbb P\) is quasismooth, and so is an orbifold, the basket \(\mathscr {B}\) is exactly the collection of quotient singularities of

*X*. Thus the numerical data \(g, \mathscr {B}\) gives the basis for a systematic search of Hilbert series with given properties, which we develop further here.

We may enumerate all Open image in new window formats Open image in new window and then list all genus–basket pairs \(g,\mathscr {B}\) whose corresponding series (9) has matching numerator. This algorithm is explained in [7, Section 4]. It works systematically through increasing \(k\in \mathbb N\), where \(k=3\bigl (\sum a_i + \sum b_i\bigr )\), the sum of the weights of the ambient space of the image of \(\mathrm{\Phi }\) in (5).

*k*is:This hints that we may have found all Fano Hilbert series that match some Open image in new window format, since the algorithm stops producing results after \(k=23\). Of course that is not a proof that there are no other cases, and we do not claim that; the results here only use the outcome of this search as their starting point, so how that outcome arises is not relevant.

### 7.2 Weighted Open image in new window varieties according to Szendrői

The elementary considerations we deploy for the key varieties Open image in new window are part of a more general approach to weighted homogeneous spaces by Grojnowski and Corti–Reid [14], with other cases developed by Qureshi and Szendrői [28, 29]. The particular case of Open image in new window was worked out detail by Szendrői [32], which we sketch here.

### Theorem 7.1

## Notes

### Acknowledgements

It is our pleasure to thank Balázs Szendrői for sharing his unpublished analysis [32] of weighted Open image in new window varieties that provided our initial motivation and the tools of Sect. 7.2, and for several helpful and informative conversations.

## References

- 1.Ahmadinezhad, H.: On pliability of del Pezzo fibrations and Cox rings. J. Reine Angew. Math.
**723**, 101–125 (2017)MathSciNetzbMATHGoogle Scholar - 2.Altınok, S., Brown, G., Reid, M.: Fano 3-folds, \(K3\) surfaces and graded rings. In: Berrick, A.J., Leung, M.C., Xu, X. (eds.) Topology and Geometry: Commemorating SISTAG. Contemporary Mathematics, vol. 314, pp. 25–53. American Mathematical Society, Providence (2002)CrossRefGoogle Scholar
- 3.Brown, G.: Graded rings and special \(K3\) surfaces. In: Bosma, W., Cannon, J. (eds.) Discovering Mathematics with Magma. Algorithms and Computation in Mathematics, vol. 19, pp. 137–159. Springer, Berlin (2006)CrossRefGoogle Scholar
- 4.Brown, G.: A database of polarized \(K3\) surfaces. Experiment. Math.
**16**(1), 7–20 (2007)Google Scholar - 5.Brown, G., Fatighenti, E.: Hodge numbers and deformations of Fano 3-folds (2017). arXiv:1707.00653
- 6.Brown, G., Kasprzyk, A.M.: The graded ring database. http://www.grdb.co.uk/
- 7.Brown, G., Kasprzyk, A.M., Zhu, L.: Gorenstein formats, canonical and Calabi–Yau threefolds (2014). arXiv:1409.4644
- 8.Brown, G., Kerber, M., Reid, M.: Fano 3-folds in codimension 4, Tom and Jerry. Part I. Compositio Math.
**148**(4), 1171–1194 (2012)MathSciNetCrossRefzbMATHGoogle Scholar - 9.Brown, G., Kerber, M., Reid, M.: Tom and Jerry: big table (2012). http://grdb.co.uk/Downloads
- 10.Brown, G., Reid, M.: Diptych varieties. I. Proc. London Math. Soc.
**107**(6), 1353–1394 (2013)MathSciNetCrossRefzbMATHGoogle Scholar - 11.Buckley, A., Reid, M., Zhou, S.: Ice cream and orbifold Riemann–Roch. Izv. Math.
**77**(3), 461–486 (2013)MathSciNetCrossRefzbMATHGoogle Scholar - 12.Chen, J.-J., Chen, J.A., Chen, M.: On quasismooth weighted complete intersections. J. Algebraic Geom.
**20**(2), 239–262 (2011)MathSciNetCrossRefzbMATHGoogle Scholar - 13.Corti, A., Pukhlikov, A., Reid, M.: Fano \(3\)-fold hypersurfaces. In: Corti, A., Reid, M. (eds.) Explicit Birational Geometry of 3-Folds. London Mathematical Society Lecture Note Series, vol. 281, pp. 175–258. Cambridge University Press, Cambridge (2000)Google Scholar
- 14.Corti, A., Reid, M.: Weighted Grassmannians. In: Beltrametti, M.C., et al. (eds.) Algebraic Geometry, pp. 141–163. de Gruyter, Berlin (2002)Google Scholar
- 15.Di Natale, C., Fatighenti, E., Fiorenza, D.: Hodge theory and deformations of affine cones of subcanonical projective varieties. J. London Math. Soc. (2017). https://doi.org/10.1112/jlms.12073, arXiv:1512.00835
- 16.Goto, S., Watanabe, K.: On graded rings. I. J. Math. Soc. Japan
**30**(2), 179–213 (1978)CrossRefzbMATHGoogle Scholar - 17.Grayson, D.R., Stillman, M.E.: Macaulay2, a software system for research in algebraic geometry. http://www.math.uiuc.edu/Macaulay2/
- 18.Iano-Fletcher, A.R.: Working with weighted complete intersections. In: Corti, A., Reid, M. (eds.) Explicit Birational Geometry of 3-Folds. London Mathematical Society Lecture Note Series, vol. 281, pp. 101–173. Cambridge University Press, Cambridge (2000)CrossRefGoogle Scholar
- 19.Ilten, N.O.: Versal deformations and local Hilbert schemes. J. Softw. Algebra Geom.
**4**, 12–16 (2012)MathSciNetCrossRefzbMATHGoogle Scholar - 20.Kawamata, Y.: Boundedness of \(Q\)-Fano threefolds. In: Bokut, L.A., Ershov, Yu.L., Kostrikin, A.I. (eds.) Proceedings of the International Conference on Algebra, Part 3. Contemporary Mathematics, vol. 131, pp. 439–445. American Mathematical Society, Providence (1992)Google Scholar
- 21.Kollár, J., Miyaoka, Y., Mori, S., Takagi, H.: Boundedness of canonical \(\mathbf{Q}\)-Fano 3-folds Proc. Japan Acad. Ser. A Math. Sci.
**76**(5), 73–77 (2000)MathSciNetzbMATHGoogle Scholar - 22.Kustin, A.R., Miller, M.: Constructing big Gorenstein ideals from small one. J. Algebra
**85**(2), 303–322 (1983)MathSciNetCrossRefzbMATHGoogle Scholar - 23.Papadakis, S.A.: Kustin–Miller unprojection with complexes. J. Algebraic Geom.
**13**(2), 249–268 (2004)MathSciNetCrossRefzbMATHGoogle Scholar - 24.Papadakis, S.A.: The equations of type \({\rm II}_2\) unprojection. J. Pure Appl. Algebra
**212**(10), 2194–2208 (2008)MathSciNetCrossRefzbMATHGoogle Scholar - 25.Papadakis, S.A., Reid, M.: Kustin–Miller unprojection without complexes. J. Algebraic Geom.
**13**(3), 563–577 (2004)MathSciNetCrossRefzbMATHGoogle Scholar - 26.Qureshi, M.I.: Constructing projective varieties in weighted flag varieties II. Math. Proc. Cambridge Philos. Soc.
**158**(2), 193–209 (2015)MathSciNetCrossRefzbMATHGoogle Scholar - 27.Qureshi, M.I.: Computing isolated orbifolds in weighted flag varieties. J. Symbolic Comput.
**79**(Part 2), 457–474 (2017)MathSciNetCrossRefzbMATHGoogle Scholar - 28.Qureshi, M.I., Szendrői, B.: Constructing projective varieties in weighted flag varieties. Bull. London Math. Soc.
**43**(4), 786–798 (2011)MathSciNetCrossRefzbMATHGoogle Scholar - 29.Qureshi, M.I., Szendrői, B.: Calabi–Yau threefolds in weighted flag varieties. Adv. High Energy Phys.
**2012**, # 547317 (2012)Google Scholar - 30.Reid, M.: Fun in codimension 4 (2011). https://homepages.warwick.ac.uk/~masda/3folds/Fun.pdf
- 31.Reid, M.: Gorenstein in codimension 4: the general structure theory. In: Kawamata, Y. (ed.) Algebraic Geometry in East Asia—Taipei 2011. Advanced Studies in Pure Mathematics, vol. 65, pp. 201–227. Mathematical Society Japan, Tokyo (2015)Google Scholar
- 32.Szendrői, B.: On weighted homogeneous varieties (2005) (Unpublished manuscript)Google Scholar
- 33.Takagi, H.: On classification of \({\mathbb{Q}}\)-Fano 3-folds of Gorenstein index 2. I, II. Nagoya Math. J.
**167**(117–155), 157–216 (2002)MathSciNetCrossRefzbMATHGoogle Scholar - 34.Taylor, R.: Fano 3-Folds and Type II Unprojection. PhD thesis, University of Warwick (In preparation)Google Scholar

## Copyright information

**Open Access**This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.